Upload app.py

app.py

@@ -0,0 +1,84 @@
+import streamlit as st
+import pandas as pd
+import numpy as np
+import xgboost as xgb
+import pickle
+import datetime
+
+# --- PAGE CONFIG ---
+st.set_page_config(page_title="Sticker Sales Predictor", layout="centered")
+
+@st.cache_resource
+def load_assets():
+    # Load the XGBoost model from the .ubj file
+    model = xgb.XGBRegressor()
+    model.load_model("model.ubj")
+    
+    # Load the exact column names used during training
+    with open("model_columns.pkl", "rb") as f:
+        model_columns = pickle.load(f)
+        
+    return model, model_columns
+
+model, model_columns = load_assets()
+
+# --- UI ELEMENTS ---
+st.title("📊 Sticker Sales Forecasting")
+st.write("Predict the number of stickers sold based on date, location, and product type.")
+
+# --- INPUT SECTION ---
+with st.container():
+    col1, col2 = st.columns(2)
+    
+    with col1:
+        date = st.date_input("Select Date", datetime.date(2026, 5, 4))
+        country = st.selectbox("Country", ["Finland", "Italy", "Norway"])
+    
+    with col2:
+        store = st.selectbox("Store", ["KaggleMart", "KaggleRama"])
+        product = st.selectbox("Product", ["Kaggle Mug", "Kaggle Hat", "Kaggle Sticker"])
+
+# --- PREPROCESSING ---
+def get_prediction():
+    # Create initial DataFrame
+    input_date = pd.to_datetime(date)
+    data = {
+        'year': [input_date.year],
+        'month': [input_date.month],
+        'day': [input_date.day],
+        'dayofweek': [input_date.dayofweek],
+        'is_weekend': [1 if input_date.dayofweek >= 5 else 0]
+    }
+    
+    df = pd.DataFrame(data)
+    
+    # Apply One-Hot Encoding manually to match training columns
+    # We create a dataframe of zeros with the same columns as the model
+    features_df = pd.DataFrame(0, index=[0], columns=model_columns)
+    
+    # Fill in the time features
+    for col in data.keys():
+        if col in features_df.columns:
+            features_df[col] = data[col]
+            
+    # Set the One-Hot features to 1
+    country_col = f"country_{country}"
+    store_col = f"store_{store}"
+    product_col = f"product_{product}"
+    
+    for col in [country_col, store_col, product_col]:
+        if col in features_df.columns:
+            features_df[col] = 1
+            
+    # Ensure column order is EXACTLY as it was in model_columns.pkl
+    features_df = features_df[model_columns]
+    
+    # Predict and inverse the log transformation (np.expm1)
+    prediction_log = model.predict(features_df)
+    return np.expm1(prediction_log)[0]
+
+# --- RESULTS ---
+if st.button("Predict Sales"):
+    result = get_prediction()
+    st.metric(label="Predicted Units Sold", value=f"{int(result)}")
+    st.balloons()